Full hardware implementation of neuromorphic visual system based on multimodal optoelectronic resistive memory arrays...

Area of Science:

• Materials Science and Engineering
• Neuroscience and Neuromorphic Computing
• Electrical Engineering and Computer Science

Background:

• In-sensor and near-sensor computing are crucial for next-generation, low-power sensory processing.
• Existing neuromorphic visual systems lack high-density, multifunctional devices for hierarchical emulation.
• There is a need for fully hardware-implemented systems with versatile image processing capabilities.

Purpose of the Study:

• To demonstrate a multimodal, multifunctional resistive random-access memory (RRAM) device array.
• To achieve a fully hardware-implemented artificial visual system with versatile image processing functions.
• To leverage modified silk fibroin protein (MSFP) for advanced neuromorphic applications.

Main Methods:

• Fabrication of a resistive random-access memory (RRAM) device array using modified silk fibroin protein (MSFP).
• Characterization of the device in both optoelectronic RRAM (ORRAM) mode (photoconductance memory) and electrical RRAM (ERRAM) mode (analogue resistive switching).
• Implementation of a full hardware artificial visual system utilizing the MSFP-based RRAM array for in-sensor and near-sensor processing.

Main Results:

• Demonstrated a novel multimodal-multifunctional RRAM device array based on MSFP.
• The device array exhibited unique negative and positive photoconductance memory (ORRAM) and analogue resistive switching (ERRAM).
• Achieved the first full hardware implementation of an artificial visual system with versatile in-sensor (image pre-processing) and near-sensor (image recognition) functions.

Conclusions:

• The MSFP-based RRAM array offers a promising platform for high-density, low-power neuromorphic visual systems.
• This technology significantly improves integration density and simplifies circuit design and fabrication complexity.
• The demonstrated system paves the way for efficient, hierarchical emulation of the retina and visual cortex.